Circulating miRNA is a useful diagnostic biomarker for nonalcoholic steatohepatitis in nonalcoholic fatty liver disease

Nonalcoholic steatohepatitis (NASH) is considered as a progressive form of nonalcoholic fatty liver disease (NAFLD). To distinguish NASH from nonalcoholic fatty liver (NAFL), we evaluated the diagnostic value of circulating miRNAs. Small RNA sequencing was performed on 12 NAFL patients and 12 NASH patients, and the miRNA expression was compared. After selecting miRNAs for the diagnosis of NASH, we analyzed the diagnostic accuracy of each miRNA and the combination of miRNAs. External validation was performed using quantitative reverse transcription PCR. Among the 2,588 miRNAs, 26 miRNAs significantly increased in the NASH group than in the NAFL group. Among the 26 elevated miRNAs in the NASH group, 8 miRNAs were selected, and in silico analysis was performed. Only four miRNAs (miR-21-5p, miR-151a-3p, miR-192-5p, and miR-4449) showed significant area under the receiver operating characteristic curve (AUC) values for NASH diagnosis. The combination of the four miRNAs showed satisfactory diagnostic accuracy for NASH (AUC 0.875; 95% CI 0.676–0.973). External validation revealed similar diagnostic accuracy for NASH (AUC 0.874; 95% CI 0.724–0.960). NASH represents significantly distinct miRNA expression profile compared with NAFL. The combination of serum circulating miRNAs can be used as a novel biomarker for the NASH diagnosis in NAFLD.

www.nature.com/scientificreports/ MicroRNA (miRNA) is a small-sized non-coding RNA comprising 20-25 nucleotides, and it binds to the target mRNA, resulting in translation inhibition or mRNA degradation 11,12 . miRNAs are synthesized in a variety of cells and participate in cell signaling 13 ; they have important roles in regulating cell growth, proliferation, and metabolism 14 . In addition to their physiological role, circulating miRNAs have been studied as a candidate for diagnosis of a variety of diseases such as malignancy and cardiovascular, neurologic, and metabolic diseases, including diabetes and NAFLD 15,16 . In patients and animal models with NAFLD, circulating miRNAs exhibit significant differences compared with healthy controls 17,18 . Circulating miRNA-34a, miRNA-122, and miRNA192 consistently increase in patients with NASH compared to patients with simple steatosis 19 . Diagnostic accuracy to distinguish NASH from NAFL was evaluated using miR-34a, but it showed moderate accuracy [area under the receiver operating characteristic curve (AUC) = 0.78)].
In the present study, we aimed to develop a biomarker to diagnose NASH in NAFLD by analyzing the circulating miRNA expression profile in sera from patients with biopsy-confirmed NAFLD using emerging next generation sequencing (NGS).

Results
Baseline characteristics. For miRNA sequencing, sera from 24 biopsy-proven patients with NAFLD were collected between February 2019 and July 2019, including 12 patients with NAFL and 12 patients with NASH. The baseline characteristics of patients are summarized in Table 1. The NASH group was more women dominant (75%) and showed lower body mass index (BMI), hemoglobin, total bilirubin, albumin, and creatinine compared with the NAFL group. Steatosis, lobular inflammation, and fibrosis did not show significant differences in severity between the NAFL and NASH groups. Ballooning is essential for the diagnosis of NASH; there was a significant difference in the presence of ballooning between the NAFL and NASH groups. The NASH group showed significantly higher NAFLD activity score (NAS) than that observed for the NAFL group. Representative histopathological findings are presented in Supplementary Fig. 1. In the validation cohort, 37 patients with biopsy-confirmed NAFLD were enrolled, including 11 and 26 patients with NAFL and NASH, respectively (Supplementary Table 1). Clinical and laboratory characteristics were similar in both groups, except for aspartate www.nature.com/scientificreports/ aminotransferase (AST). We observed no significant differences for steatosis, lobular inflammation, and fibrosis between the NAFL and NASH groups in the validation cohort; however, hepatocyte ballooning and NAS were found to be significantly higher in the NASH group compared with those in the NAFL group in the validation cohort.
miRNA expression profiles. In miRNA analysis, a total of 2588 miRNAs were analyzed, and each sample expressed 332 to 618 miRNAs (median 469 miRNAs). Approximately 2153 miRNAs were excluded from analysis because these miRNAs were expressed in < 50% samples; therefore, 435 miRNAs were compared between NAFL and NASH groups ( Supplementary Fig. 2). A total of 38 miRNAs showed a significant difference in expression between the NAFL group and NASH group as shown in the heatmap (Fig. 1A). Twenty-six miRNAs significantly increased more than two times in the NASH group as compared to the NAFL group (P < 0.05) (Fig. 1B). In contrast, 12 miRNAs significantly decreased in the NASH group as compared to the NAFL group (P < 0.05).
We also compared miRNA expression between patients with and without fibrosis to identify the confounding effect of fibrosis in NASH ( Supplementary Fig. 4). Although 14 miRNAs showed significant differences in expression, there was no overlap between these 14 miRNAs and 8 miRNAs that were selected to evaluate the diagnostic accuracy for NASH in Supplementary Fig. 2.

Discussion
As the need for noninvasive testing to determine the severity of NAFLD increases, various biomarkers have been investigated to diagnose NASH or advanced fibrosis. However, biomarkers for NASH that represent progressive inflammation of hepatocytes have shown limited accuracy and accessibility 7 . In this study, we found a combination of several circulating miRNAs using NGS and qRT-PCR, which could be a useful biomarker to diagnose NASH. miRNA is a small non-coding RNA, comprising ≤ 25 nucleotides. The major role of miRNAs is post-transcriptional inhibition of gene expression by binding to the 3′-untranslated region of target mRNAs 23 . Circulating miRNAs are derived from specific cells and reflect the presence of disease or disease severity. The expression of miRNAs depends on the presence of NAFLD or the disease severity of NAFLD; therefore, miRNA expression has been suggested as a diagnostic biomarker for NAFLD, NASH, and advanced fibrosis 19 . Several circulating miRNAs, such as miR-16, miR-21-5p, miR-34a, miR-122, miR-192, and miR-375, have shown higher expression in the sera of patients with NASH than in those with NAFL. Most of these studies evaluated miRNA expression using real-time PCR quantification. More than 2600 sequences of human mature miRNAs are known in miRBase 24 ; therefore, there is a limitation in evaluating the overall expression profile of whole human mature miRNAs. Our study analyzed a total of 2588 mature circulating miRNA expression profile using NGS. We selected four candidate miRNAs as biomarkers for NASH diagnosis and analyzed their diagnostic value using normalized values from NGS data. A combination of four miRNAs, miR-21-5p, miR-151a-3p, miR-192-5p, and miR-4449 showed significant diagnostic accuracy with an AUC of 0.875 in the normalized value from NGS data. According to external validation using qRT-PCR, we found that the diagnostic accuracy of a combination of miR-21-5p, miR-151a-3p, miR-192-5p, and miR-4449 was sufficient for them to be a biomarker for NASH diagnosis. Among human mature miRNAs, miR-122-5p is the major miRNA that is expressed in the liver, and circulating miR-122-5p increases in patients with NAFLD 25 . Although miR-122-5p was the second most abundant circulating miRNA in our study after miR-423-5p, the difference in expression level of circulating miR-122-5p between NAFL and NASH groups was not significant. miR-21-5p increases in the plasma of patients with NASH 26 and it is associated with hepatic metabolism, inflammation, and lipid metabolism 27,28 . miR-21-5p has been identified as a typical onco-miRNA in many previous studies. Our study indicated that miR-21-5p was highly expressed in NASH, and the relationship among miR-21-5p, NASH, and liver cancer could be an interesting topic for further study. Plasma level of miR-151a-3p is positively correlated with that of TNF-α, which is the classical inflammatory parameter and major factor in the progression of NAFLD 29 . Circulating miR-192-5p is also upregulated in patients with NASH as compared to patients with NAFL 17 . Exosomes from lipotoxic hepatocytes showed increased miR-192-5p, and exosomal miR-192-5p regulates disease progression of NAFLD by activating proinflammatory macrophages 30,31 . miR-4449 expression is rarely known in patients with NAFLD. In patients with obesity, circulating exosomal miR-4449 showed increased expression as compared to the healthy control group, and its expression was decreased after bariatric surgery 32 . The increasing patterns of our miRNAs exhibited consistency with other liver-related pathology states or obesity.
We applied a previous study uploaded in GEO to network analysis. From the GSE48452 dataset, we identified 265 downregulated genes in patients with NASH, and 26 pairs of miRNA-mRNA interactions were selected. Thus, bioinformatic analysis was applied to explore the correlation between miRNA-mRNA expression profiles. The target genes downregulated with upregulated miRNAs in NASH could be revealed by the public expression dataset.
This study has some limitations. First, the expression levels of miRNAs were standardized to U6 as an internal control. Although not all studies used snRNA U6 as internal control, many other studies used snRNA U6 as internal control [33][34][35] . Further validation studies are required to evaluate the possible clinical application of miRNAs as diagnostic biomarkers using other standardized control. Second, sequencing was conducted using a small number of patients. To overcome this limitation, we conducted external validation with patients from other centers, and we found similar diagnostic accuracy for NASH in an external validation cohort. However, further validation studies with larger populations and varying degrees of fibrosis are required. Third, we could only provide relative expression level of miRNA for the diagnosis of NASH in patients with NAFLD not absolute expression level. Further studies are required for clinical application.
In conclusion, NASH represents significant distinct miRNA expression profiles compared with NAFL. A combination of serum circulating miRNAs including miR-21-5p, miR-151a-3p, miR-192-5p, and miR-4449 could be used as a novel biomarker for the diagnosis of NASH in NAFLD.

Patients and methods. Patients and sera collection.
For small RNA sequencing, 24 patients with biopsyconfirmed NAFLD, comprising 12 NAFL patients and 12 NASH patients, were enrolled from our previous study at Korea University Guro Hospital 36 . Another 37 patients with biopsy-confirmed NAFLD from Anam Hospital were enrolled for comparison of miRNA expression between NAFL and NASH in external validation. Patients were excluded if they had consumed excessive alcohol and had viral hepatitis, autoimmune hepatitis, primary biliary cholangitis, decompensated cirrhosis, or other severe systemic diseases.
Laboratory tests were performed before liver biopsy, and sera were collected during blood sampling for laboratory tests. Sera were stored at -80℃ and thawed just before RNA extraction. This study had been approved by the institutional review board from Korea University Guro Hospital (2016GR0302) and Anam Hospital (2018AN0129). All patients agreed to the sera collection and submitted written informed consent. All investigators conducted this study in accordance with the Declaration of Helsinki. Each author reviewed and approved the final manuscript. Quantitative reverse transcription PCR. cDNA synthesis was done using reverse transcriptase with miRNAspecific stem-loop primers (Applied Biosystems). qRT-PCR was performed on a QuantStudio 6 Flex Real-time PCR system (Applied Biosystems) using Taqman master mixture. The relative abundance of miRNA was normalized to that of small nuclear RNA U6. The relative amount of each miRNA was measured using the 2 −∆∆Ct method. The primers are summarized (Supplementary Table 3).
mRNA data collection and analysis. Data mining from Gene Expression Omnibus database (GEO; https:// www. ncbi. nlm. niste atohe patit ish. gov/ geo/) was performed to confirm a spectrum of differentially expressed mRNA profiles of NASH. Gene microarray expression profiles between the NAFL group and NASH group were collected from GEO using the keyword "steatosis and steatohepatitis. " GSE48452 was suitable for our analysis and downloaded to select genes differently expressed in the NASH group. Differently expressed genes were retrieved by t-test using R software and filtered when log 2 (fold change) > 1 or < − 1, and P < 0.05 between the NAFL group and NASH group. An miRNA-mRNA network was constructed using mirDIP 21